Learn
Getting paid to train AI, explained.
People now earn money teaching AI systems: rating answers, labeling data, lending expertise, recording their voice. Most have never heard the job titles. This is the plain-language map: what each kind of work actually is, what you would do day to day, who it suits, and how to start. When you want the numbers, see what each one pays.
Pay figures come from live open roles wherever enough of them publish a rate, and from the compiled benchmark otherwiseupdated 9 hrs ago
The map
Three families of AI training work.
Every kind of AI training work falls into one of three groups, sorted by what you bring to it. Find the family that fits your background, then open a guide to see if the work is for you.
Evaluate & give feedback
Judge and stress-test what models produce.
You read what an AI produced and decide how good it is: rate it, rank it against other answers, or actively try to make it fail. This is the most common on-ramp into AI training work, and it rewards careful judgment and clear writing far more than any technical background.
Rate, rank, and compare AI answers so a model learns which response is better.
Red-Teaming & AI Safety
MediumGet paid to break AI models: find the unsafe, biased, or wrong answers before the public does.
Get paid to take part in surveys, experiments, and human-preference studies that feed AI research.
Contribute expertise
Bring specialist knowledge the model needs.
You use real expertise (a language you program in, an advanced degree, a professional license) to write and verify content most people cannot. This is the highest-paid family, gated by genuine skill or credentials rather than by volume. It now includes the 2026 frontier: designing the interactive tasks and reinforcement-learning environments that agentic models train against.
Review, write, and grade real code so AI models learn to program correctly.
Math & STEM
HighWrite and check hard math, physics, chemistry, and biology problems (including formal proofs) that teach and test AI.
Use your law, medicine, or finance credential to judge whether AI reasoning is actually correct.
Build and grade the interactive tasks and RL environments that agentic AI trains against.
Create & label data
Produce or annotate the raw material models learn from.
You make the training data itself: labeling images, recording or transcribing speech, supplying native-language data, or writing reference answers. This is the most accessible family, with the widest range of tasks available worldwide.
Label images, video, LiDAR, and text so AI models can perceive and categorize the world.
Give AI models native-language data, translation, and dialect judgment humans cannot fake.
Record, transcribe, and label voice audio so AI can understand and produce human speech.
Write prompts, model answers, and edits that teach AI to write clearly and well.
Guides
The hard parts, in depth.
What is an AI trainer, and how do you become one
What an AI trainer does, what the work pays by kind, what roles require and how applications work, from our live job data, including how to start without experience.
36 min read
AI interviews: how they work and how to pass one
How AI interviews at Mercor, micro1 and Braintrust work, what the AI scores, how to prepare and answer, retake rules, your rights, and what happens after.
48 min read
The fastest-growing jobs in 2026: the top 10, the top 100 and the jobs in decline
The 10 job categories adding the most jobs and the most value in 2026, the jobs shrinking fastest worldwide, and the 100 fastest-growing jobs, with numbers.
63 min read
Free tools
Check an offer. Practise the interview.
Not sure where to start?
Pick by what you already have.
You just want to start earning
Begin with the lowest-barrier work: rating AI answers, research studies, or data labeling. No credentials, quick to start.
Start with RLHF & evaluation →You have a profession or degree
Your expertise is the premium. Doctors, lawyers, engineers, and PhDs earn the highest rates verifying model reasoning in their field.
See expert work →You speak another language
Native fluency in an in-demand language is a real, paid skill. Multilingual and audio work is open worldwide.
See multilingual work →